@agentcommunity_

agents need names.

Joined January 2025
On 12 August at 23:19 UTC, forty minutes before ICANN's application window closed, we filed the Community application for .agent. We did it. As promised. You cannot apply for a new top-level domain whenever you want. ICANN last opened this window in 2012, and nobody knows when the next one comes. This round closed with 1,600+ applications. One of them is ours. It was carried by a community of nearly 30,000 builders, researchers, and organizations. A majority of members signed letters endorsing the application - Brave, Hugging Face, Replit, Product Hunt, NEAR AI among them, and many, many more. Why do this? Agents are starting to act on people's behalf. When an agent books a flight or signs a contract, the question that matters is simple: who answers for what it just did? A .agent name, if the TLD is delegated, is proposed to make that answer checkable by default. Not a badge you can buy. What it took: - 225 questions. - Months of writing and rewriting. - Five hours just to submit. And the biggest lesson of the whole process? ICANN does not score hopes and believes. The policies we filed are proposed to become binding contract, enforceable by ICANN Compliance. Musts and not just mays The evaluation is still ahead of us and the decisions rest with ICANN. We won't predict anything. And while it is only the beginning of the future, it was a massive milestone. The full story, learnings included: agentcommunity.org/blog/we-f…
2
2
9
806
Models will replace large swaths of Playwright tests by 2027 by crawling apps through desktop usage, per @ThePrimeagen.
2027 prediction: Models will be replacing tons of playwrite tests as a means to crawl and use your application using desktop usage. Just doing it now for Omarchy just shows how crazy powerful it is and how hard it is to do the same thing with scripts.
1
32
Dan Kornas surfaced roam-code, a local codebase-intelligence CLI and MCP server indexing repos into a SQLite code graph for blast-radius assessment before edits.
AI coding agents shouldn’t guess what an edit will break. roam-code is a local codebase intelligence CLI and MCP server for builders using AI coding agents. It helps you check a change before you make it by indexing your repository into a SQLite-backed code graph, then surfacing dependencies, affected tests, complexity, and architectural context. Key features: • Preflight checks – assess a symbol’s blast radius, affected tests, complexity, and architecture rules before editing • Local code graph – index symbols, calls, imports, Git history, and other codebase relationships across 28 languages • Health analysis – inspect codebase health, cycles, coupling, dead code, and related structural findings • MCP server – connect Roam’s graph-aware queries to MCP-aware coding tools through selectable tool presets • Change evidence – generate tamper-evident ChangeEvidence packets to record what was checked around a change It’s open-source (Apache License 2.0). Link in the reply 👇
1
44
@davis7 warned swarms are not worth it for most tasks and will obliterate usage, recommending them only for really big complex one-shot work.
Replying to @rohit3a @executorsh
nah they're not worth it for most things, message boards and big swarms are not worth it outside of really big complex work (one shot massive apps, giant 3d jobs, puzzles, etc.). Any problem that really needs to be "burned down" It will obliterate your usage, esp since it's only really worth on max reasoning (ultra)
36